> -----Original Message----- > From: Liu, Hongtao > Sent: Monday, August 31, 2026 1:03 PM > To: H.J. Lu <[email protected]>; GCC Patches <[email protected]>; > Uros Bizjak <[email protected]> > Subject: RE: [PATCH] x86: Compile vect-unpack-[123].c with -mtune=generic > > > > > -----Original Message----- > > From: H.J. Lu <[email protected]> > > Sent: Monday, August 31, 2026 7:41 AM > > To: GCC Patches <[email protected]>; Uros Bizjak > > <[email protected]>; Liu, Hongtao <[email protected]> > > Subject: [PATCH] x86: Compile vect-unpack-[123].c with -mtune=generic > > > > Compile vect-unpack-[123].c with -mtune=generic to avoid > > > > FAIL: gcc.target/i386/vect-unpack-1.c scan-assembler-times vpmovzxbw[ > > \\t]+[^\n]*%zmm 2 > > FAIL: gcc.target/i386/vect-unpack-2.c scan-assembler-times vpmovsxbw[ > > \\t]+[^\n]*%zmm 2 > > FAIL: gcc.target/i386/vect-unpack-3.c scan-assembler-not vpmovsxbw[ > > \\t]+[^\n]*%zmm > > > > when different tunings are used. > Use -mprefer-vector-width=512?
Hmm, for znver5, there's masked epilogue, so it still failed with -mprefer-vector-width=512. Ok with the original patch(generic tune applied). > > > > > PR testsuite/126923 > > * gcc.target/i386/vect-unpack-1.c (dg-options): Add -mtune=generic. > > * gcc.target/i386/vect-unpack-2.c (dg-options): Likewise. > > * gcc.target/i386/vect-unpack-3.c (dg-options): Likewise. > > -- > > H.J.
